AI for Cybersecurity in Distributed Automotive IoT Systems
Published in Springer Nature, 2023
Recommended citation: V. K. Kukkala, S. V. Thiruloga, and S. Pasricha, "AI for Cybersecurity in Distributed Automotive IoT Systems," in Frontiers of Quality Electronic Design (QED), Springer Nature, 2023.
Abstract
The widespread adoption of autonomy in vehicle subsystems has led to modern vehicles being highly connected to various external systems, to achieve robust and high efficiency driving performance. This trend has led to increased software and hardware complexity in distributed electronic control units (ECUs) inside vehicles, as well as an increase in ECU count. Modern automotive systems thus represent highly complex distributed cyber-physical IoT system. Unfortunately, the increased system complexity and interactions with external systems have made the automotive systems highly vulnerable to various cyberattacks. The growing compute power and the large availability of automotive network data make emerging artificial intelligence (AI)-based solutions a promising choice for improving cybersecurity in automotive systems. In this chapter, we provide an overview of novel AI-based intrusion detection system (IDS) framework called INDRA that actively monitors messages in the in-vehicle network to detect cyberattacks in automotive systems. INDRA utilizes a gated recurrent unit (GRU)-based recurrent autoencoder architecture which helped in achieving a superior IDS performance with low overhead (low memory footprint and faster detection times). The evaluation of INDRA under various attack scenarios, and comparison with the state-of-the-art prior works is presented at the end of the chapter.